Pediatric arteriovenous malformations (pAVMs) are a challenging entity in which hemorrhagic presentation occurs in more than half of cases.^1–7 There is little consensus regarding the management of pAVMs beyond initial stabilization following rupture.^2–6,8 Observation, surgical removal, stereotactic radiosurgery, embolization, or a combination of these options have often been chosen at least partially based on the availability of interventions and the preferences of the practitioners involved. The availability of Gamma Knife radiosurgery (GKRS) as an adjunct or alternative to resection or embolization has had substantial effects on the management of pAVMs, particularly in cases in which the potential for operative morbidity is excessive. Our center has a more than 3-decade history of multimodality management of AVMs in which all of the current management options have been used.⁹ We sought to characterize the long-term outcomes of our cohort in the multimodality era.

Methods

We conducted a retrospective cohort study of pAVM outcomes in patients cared for at the University of Pittsburgh and Children’s Hospital of Pittsburgh during a 30-year interval (January 1, 1988–December 31, 2018). Retrospective review of clinical data was approved under the University of Pittsburgh Medical Center Institutional Review Board guidelines. The presenting symptoms, long-term modified Rankin Scale (mRS) outcome score, patient demographic information, management strategies, angioarchitecture and other radiographic characteristics, and long-term outcomes were reviewed. Recurrences and post-treatment hemorrhages were documented.¹⁰

Results

Of the 212 patients, 31 (14.6%) had an mRS score of ≥ 3 on most recent follow-up, and 181 (85.4%) had an mRS score of ≤ 2 on most recent follow-up. One hundred nine patients (51.4%) were male and 103 (48.6%) were female, with 17 male and 14 female patients having poor outcomes (p = 0.68). The mean age on presentation was 11.3 years, with the mean age of patients with good outcome being 11.6 years versus 9.6 years in patients with poor outcome (p = 0.015). The median mRS score on presentation for the entire cohort was 1, with a median presenting mRS score of 1 for patients with good long-term outcomes, and a median presenting mRS score of 3 for patients with poor long-term outcomes (p < 0.0001) (Table 1). Seventeen patients (8%) were asymptomatic—their lesions were discovered during imaging for other diagnoses—and 195 (92%) were symptomatic. Overall, patients with motor deficits and obtundation were more likely to have a poor outcome, whereas asymptomatic patients and patients presenting with seizures, headaches, altered mental status, and visual field cuts were more likely to have a good long-term outcome (p < 0.0001). One hundred twenty-eight patients (60.4%) had pAVM-associated hemorrhage on presentation, with 102 (56.4%) of 181 patients in the good outcome group and 26 (83.9%) of 31 patients in the poor outcome group presenting with hemorrhage (p = 0.01). The 5 patients without hemorrhage who had poor outcome included 1 patient who developed progressive epilepsy and hemiparesis from pAVM mass effect, 1 patient whose family refused treatment on incidental presentation and who suffered a fatal hemorrhage subsequently, 1 patient who had persistent deficits from brainstem mass effect even with GKRS, and 2 patients who suffered hemorrhage years after treatment. Fifty (73.5%) of 68 patients less than 10 years of age had a hemorrhagic presentation (p = 0.0067) (Table 2). The mean follow-up was 79.7 ± 62.1 months. Only 13 (6.1%) patients had less than 1 year of clinical follow-up, of whom 2 died.

Spetzler-Martin Grade, Treatment Type, and Long-Term Outcomes

When stratifying by Spetzler-Martin grade, patients who had a grade of I–III and who had a craniotomy as part of their care had a mean mRS score of 1.2 on long-term follow-up, with 38 (81%) of the 47 patients having a good outcome. In contrast, patients who underwent GKRS alone with a grade I–III pAVM had a mean mRS score of 0.7 on long-term follow-up, with 83 (93%) of 89 patients having a good outcome (p = 0.028). However, when considering patients who underwent resection without a hemorrhagic presentation, the mean mRS score on follow-up was 0.55 versus a mean of 1.4 among those with a hemorrhagic presentation (p = 0.002). Patients with a Spetzler-Martin grade IV–V pAVM who underwent a craniotomy as part of their care had a mean mRS score of 3.7, with 2 (29%) of 7 patients having a good outcome. In contrast, patients who underwent GKRS alone with a grade IV–V pAVM had a mean mRS of 1.1 on long-term follow-up, with 35 (85%) of 41 patients having a good outcome (p = 0.0044). All of the patients who underwent craniotomy with a high-grade AVM presented with hemorrhage requiring evacuation. Thirteen patients who underwent craniotomy initially for decompression of hemorrhage subsequently had other treatment. When excluding these patients, the mean long-term mRS score for the remaining patients with craniotomy was 1.1 (Spetzler-Martin grade I–III: mean mRS 1, Spetzler-Martin grade IV–V: mean mRS 2) versus 2.8 (Spetzler-Martin grade I–III: mean mRS 1.9, Spetzler-Martin grade IV–V: mean mRS 4.6) for the 13 excluded patients (p = 0.002) (Fig. 1). Excluding these 13 patients, there was no difference in outcomes between craniotomy and GKRS for any Spetzler-Martin grade (p = 0.37).

Comparison of good (mRS < 3) and poor (mRS ≥ 3) outcomes by treatment type. The p values represent comparison of the designated group to the remainder of the cohort except where otherwise indicated.

Discussion

In this report we demonstrate the following: 1) the multimodality management of pediatric pAVMs is associated with a high rate of long-term functional independence and obliteration; 2) recurrences, while rare, frequently occur many years after obliteration; and 3) the strategic application of different treatment modalities is not independently associated with a better or worse outcome when applied to individual patient circumstances such as hemorrhagic presentation or high Spetzler-Martin grade lesions.

Eighty-five percent of our cohort were functionally independent on long-term follow-up, with a total of 6 deaths, which reflects a superior outcome compared to older cohorts but is comparable to more recent ones.^1,11,12 Of the deaths, 3 were directly related to brain hemorrhage from pAVMs, 2 were associated with unrelated medical conditions that developed after pAVM treatment, and 1 was due to a hematological malignancy. Overall, no differences in outcome based on treatment modality were identified. However, this was in part due to selective application and combination of techniques based on individual cases. GKRS was found to be effective in the management of many low and high Spetzler-Martin grade lesions, with craniotomies being predominantly reserved for low-grade lesions and emergency decompression. As such, patients with high-grade lesions are routinely advised to pursue nonresective treatment, whereas situations in which the patient is unstable due to hemorrhagic presentation are routinely managed with craniotomy. Families of patients with low-grade lesions are offered both craniotomy and GKRS options, with deference to personal preference in situations of equipoise. High-density niduses are believed to be particularly ideal for GKRS. Embolization has gradually become seen as a secondary option in situations in which deep arterial feeders can be accessed in order to reduce operative complexity and in situations in which high-risk features such as arterial aneurysms are noted. Overall, however, patients are provided the opportunity to go forward in life to achieve their potential as we hope will happen for all of our patients (Fig. 2).

Artistic rendition of pAVMs by a former patient presenting with hemorrhagic pAVM, treated by our center, who went on to become a professional graphic designer and artist in adulthood. Copyright Michael M. McDowell and Stephanie Greene. Published with permission. Figure is available in color online only.

In our study, approximately 80% of patients with radiographic follow-up demonstrated obliteration. Most obliteration rates in the pAVM literature hover around 60%–70%.^12–14 However, substantial differences in the definition of obliteration and the accepted radiographic modalities exist. The majority of our patients had a documented formal angiogram after treatment. It is our general practice to rely on MRA for post-treatment screening purposes in order to minimize the radiation burden, but a formal angiogram after treatment is recommended. Typically, we follow patients who undergo GKRS for 3 years with MRA. At this point, if the AVM is no longer noted on MRA, a formal angiogram is used to confirm this. If residual pAVM is noted, an angiogram may be performed as part of additional treatments. For patients who underwent other treatment modalities, an angiogram 6–12 months after treatment is recommended. For all patients, MRA annually is typically recommended, although for patients with exceptionally long follow-up this may be gradually spaced out to every 3 years. Of note, the mean dose for GKRS was 20 Gy in this series. In a recent study, Starke et al. demonstrated that marginal GKRS doses of 22 Gy or higher appeared to be associated with higher rates of obliteration.⁸ In-field obliteration has been correlated to marginal dosing, but this may in part be secondary to higher doses reducing the volumetric inaccuracy of AVM nidus localization during angiogram-based radiosurgery.^15,16

Recurrence separate from residual pAVM was noted in 5 cases during follow-up. Recurrence was frequently found many years after documented obliteration and may potentially increase in frequency during the lifespan of this cohort. We recommend continued imaging follow-up beyond obliteration every 3–5 years in pediatric patients. In most cases, MRA is the most appropriate modality for screening, with formal angiography reserved for concerning findings. Recurrence was found to be associated with craniotomies without GKRS in the context of hemorrhagic presentation. Microsurgery may not allow for the visualization of all aspects of abnormal microvasculature during resection in a population with potentially higher capacity for vascular neogenesis and thus regrowth of pAVMs. Hemorrhage may reflect a higher concentration of local growth factors, prompting the development of friable vessels prone to bleeding and also resulting in recurrence despite complete resection.

Our overall hemorrhage rate was consistent with past literature. Hemorrhage was not found to be independently associated with poor outcome.^{1,3,4,11,17,18} This should come with the caveat that multivariate analysis does not provide insight into causality, merely the strength of association. Of patients presenting without hemorrhage, only 3 of 84 presented with an mRS score > 2, versus 62 of the 128 patients presenting with hemorrhage (p < 0.0001). However, although hemorrhage was strongly correlated with a higher mRS score, a hemorrhagic presentation without an initially elevated mRS severe disability was unlikely to result in a poor outcome. Hemorrhage also was associated on univariate analysis with poor outcomes in children under the age of 10 years, in part due to 4 of the 6 deaths occurring in younger children. This group also had a higher proportion of hemorrhagic presentation in comparison to older patients, potentially in part due to a lower tendency to notice subtle neurological signs (such as headaches) in very young children. In addition, unstable AVMs may be more likely to present at an early age due to a greater bleeding tendency. Ultimately, patients with hemorrhage frequently benefit from craniotomy for both decompressive and therapeutic reasons. However, these patients may also be particularly vulnerable to having residual or recurrent pAVMs, in which case GKRS should be considered as an alternative to repeat craniotomy. Craniotomies instead of GKRS for definitive resection of low Spetzler-Martin grade, unruptured lesions remain a viable option and should be presented to families during discussion of treatment options, particularly in patients with angiographic features concerning for hemorrhage risk, such as aneurysms.

On multivariate analysis, post-GKRS hemorrhage trended toward an association with poor outcome. Post-GKRS hemorrhage is relatively uncommon, but has been documented in a large multicenter database in up to 9% of pediatric cases.¹⁷ In our cohort, hemorrhage occurred in 5.5% of patients who underwent GKRS, with the mean time until hemorrhage being 2 years after initial treatment. Post-GKRS hemorrhage warrants counseling in cases in which therapeutic equipoise may exist. Nonetheless, our data demonstrate that GKRS overall is a valuable and safe treatment option, particularly in high-grade lesions in which surgical and endovascular management are not tenable.

Conclusions

High rates of long-term functional independence can be achieved with comprehensive multimodality treatment of pAVMs. At our center, we did not find a difference in outcome based on treatment choice when accounting for factors such as Spetzler-Martin grade and presenting morbidity from hemorrhage. Recurrences, although infrequent, typically occur years after obliteration, and thus “cured” patients warrant long-term screening.

Acknowledgments

We thank Alessandra Crivelli, a pAVM survivor, for her artwork contribution.

Disclosures

Dr. Agarwal receives royalties from Thieme Medical Publishers. Dr. Lunsford is a consultant for Insightec and DSMB, and he has direct stock ownership in Elekta.

Author Contributions

Conception and design: McDowell, Agarwal, Greene. Acquisition of data: McDowell, Agarwal, Mao, Johnson, Kano, Greene. Analysis and interpretation of data: McDowell, Agarwal, Mao, Johnson, Kano, Greene. Drafting the article: all authors. Critically revising the article: all authors. Reviewed submitted version of manuscript: all authors. Approved the final version of the manuscript on behalf of all authors: McDowell. Statistical analysis: McDowell.